From better schools to better nourishment: evidence from a school-building program in India

	Mean	Standard deviation

	(1)	(2)
BMI	18.79	13.22
BMI (if underweight)	15.43	1.94
Age 11–22 years (=1, if yes)	0.51	0.49
SC/ST (=1, if yes)	0.74	0.44
Female (=1, if yes)	0.49	0.49
Age (in years)	17.68	7.08
Education (among males; highest number of years)	7.97	4.90
Education (among females; highest number of years)	5.66	5.21

BMI density by KGBV cohorts.
*Notes*: The figure plots the density of BMI values. The area between the two vertical lines is the potential healthy zone, i.e., BMI between 18.5 and 25. The panel on the right plots the densities for the affected cohort, i.e., girls of lower castes in the affected age cohort, and the left-hand side panel includes everyone else.

Table 2

Probability of being in the healthy BMI zone

	KGBV	Non-KGBV	Δ

	(1)	(2)	(1)–(2)
Healthy BMI	0.3088	0.2311	0.0777
Standard deviation	0.0034	0.0015	H₀: Δ = 0
			p-value < 0.001

Table 3

Summary of identification strategy

Age in data set	6–10 years	11 12 13 14	15 16 17 18 19 20 21	22 and above
Age in policy year	0–3 years	4 5 6 7	8 9 10 11 12 13 14	15 and above
Exposure to policy	Not yet exposed	Currently exposed	Previously exposed	Never exposed

Table 4

Effects of KGBV on health status: triple-difference methodology

	Extensive margin	Body mass index (if BMI < 18.5)
	BMI < 18.5	KGBV effect	Falsification

	(1)	(2)	(3)
KGBV	–0.0043	0.1901**	0.0742
	(0.0125)	(0.0922)	(0.1902)
R²	0.33	0.41	0.45
Observations	90,988	33,406	16,756
Mean of dependent variable	0.37	15.43	15.10

[i] Notes: Column 1 suggests no effects of the program at the extensive margins of health; Column 2 presents the results on the intensive margin effects, whereas Column 3 reports the falsification results at the intensive margin. Therefore, the sample in columns 1–2 is restricted to only individuals in the age group of 6–30 years in the IHDS-II data set (2011–12), which is a period after the KGBV was implemented. In Column 3, the sample is from the pre-policy period, i.e., IHDS-I. All the columns report results from different regressions. Column 1 is a regression on the full sample. Columns 2 and 3 are the results from restricted subsamples. For columns 2 and 3, the sample is restricted to only the low-BMI individuals, with BMI <18.5. It only makes sense for an increase in BMI for this subcategory of the population. The extensive margin of this measure is essentially the outcome variable in Column 1. The coefficient KGBV is the causal effect of the KGBV program on outcomes, as described in the section on the triple-difference estimation strategy. All regressions include the regional (village) fixed effects and control for the relevant baseline dummy variables and double interactions. Additional controls are age, family size, and education of household head, both female and male. Robust standard errors clustered at the regional (village) level are reported in parentheses. **p < 0.05.

Table 5

Cohort convergences: difference-in-differences

	Body mass index (if BMI < 18.5)
	Females	Disadvantaged castes

	(1)	(2)
affected*disadvantaged	0.1984***
	(0.0716)
girl*affected		0.3916***
		(0.0427)
R²	0.47	0.44
Observations	17,177	25,799
Mean of dependent variable	15.56	15.41

[i] Notes: Sample includes 6- to 30-year–old individuals in IHDS-II with BMI values <18.5. In Column 1, the subsample is restricted to only females. The estimated coefficient gives the effect of the intervention on the affected cohort’s (disadvantaged groups) BMI compared to other groups. In Column 2, the subsample is restricted to only the disadvantaged groups. So, the estimated coefficient gives similar cohort effects for girls relative to boys, only in this subcategory. All regressions include regional (village) fixed effects and controls for the relevant baseline dummy variables. Additional controls are age, family size, and education of household head, both female and male. Robust standard errors clustered at the regional (village) level are reported in parentheses. ***p < 0.01.

Triple differences along gender and disadvantaged dummies by age.

Table 6

Robustness check: restrictive subsamples

	Body mass index
	Age < 30	Age < 25	Age < 22
	(1)	(2)	(3)
KGBV	0.1901**	0.1832*	0.2131**
	(0.0922)	(0.0951)	(0.0985)
R²	0.41	0.41	0.39
Observations	33,406	30,723	29,389

[i] Notes: All columns report results from different regressions. The coefficient KGBV is the causal effect of the KGBV program on outcomes, as described in the section on estimation strategy. All regressions include regional (village) fixed effects and controls for relevant baseline dummy variables and double interactions. Additional controls are age, family size, and education of household head, both female and male. Robust standard errors clustered at the regional (village) level are reported in parentheses. **p < 0.05; *p < 0.1.